A Micro-expression Recognition Method Based on Optical Flow and Fisher Vector Coding

Abstract

A micro-expression recognition method based on optical flow and Fisher Vector encoding is disclosed, including: performing HOF feature extraction on the micro-expression image sequence in the test sample, and obtaining the first feature data X={x 1 ,...x r ,...x R |x r ∈R D}; Perform Fisher Vector encoding on the first feature data X to obtain second feature data F; identify the second feature data F based on the pre-trained classifier. The present invention uses optical flow features to describe the weak movement of micro-expression, and uses Fisher Vector to encode optical flow features, and further extracts expected information and variance information, so it can more fully represent micro-expression features and reduce the influence of local noise , which improves the robustness of micro-expression features, and then improves the recognition rate of micro-expression recognition.

Description

technical field [0001] The invention relates to the technical field of machine learning and pattern recognition, in particular to a micro-expression recognition method based on optical flow and FisherVector coding. Background technique [0002] The study of human facial expressions originated in the 19th century. Recently, foreign scholars Ekman and Erika conducted a study on facial mapping behavior, and verified that micro expressions can provide more comprehensive covert emotional information. Different from macro-expression, micro-expression is a kind of rapid expression with a short duration and low intensity. Although the duration is short, micro-expressions can better reveal the true emotions in people's hearts and provide a reliable basis for judging people's inner mental state. Therefore, they have important applications in the judicial system and clinical diagnosis. [0003] In foreign countries, the study of micro-expressions started earlier, not only proposing a...

AI Technical Summary

Problems solved by technology

[0004] However, the existing micro-expression recognition methods have problems such as poor robustness, easy to be disturbed by noise, and low recognition rate.

Images